EV Fleet Transition: Which Vans to Switch First? Model It With AI

Start by scoring each van on four things: how far it drives on its longest realistic day, where it sleeps and whether it can charge there, how many miles it covers a year, and when it is due for replacement. Switch first the vans that pass all four. AI builds the spreadsheet; you supply and check the numbers.

The ranking that comes out is rarely the one people expect. The newest diesel van is not necessarily the best candidate. A tired van with 30,000 miles a year and a depot charger often is. A van parked outside the driver's home with no charging point is a poor candidate however short its route.

Follow me on Instagram@sagnikteaches

The four filters, in the order that rules vans out

Work through them in this order, because each cheap filter saves you modelling the vans that cannot pass it.

Connect on LinkedInSagnik Bhattacharya

1. Longest realistic day. Take the longest day in the last six months that happens at least once a month, not the average. If it fits within about 70 per cent of the van's real-world range (a margin for cold weather, load and battery ageing), the van passes. Otherwise it needs a mid-day charge or stays diesel.
2. Overnight charging. Can the van plug in where it sleeps, at a depot or at a driver's home with a safe, paid-for arrangement? Relying on public rapid chargers for everyday use is the costliest way to run an electric van, so a van with no overnight charging usually waits.
3. Annual mileage. Electric vans cost more to buy and less to run. The more miles a van does, the faster the running saving repays the extra price.
4. Replacement timing. Switching a van that has three years of useful life left throws away value. Prefer vans due for replacement within the next 6 to 18 months.

The order matters because the first two are pass or fail, while the second two are about money. A van that fails filter one or two does not need a cost model yet. That also keeps the spreadsheet small enough to check by hand, which is the point of building it.

Subscribe on YouTube@sagnikteaches

Gather the route and parking facts first

Most of the effort is data collection, not modelling. For each van you need:

• Daily distances for at least 90 days, from a telematics export, fuel-card statements or the drivers' mileage logs. A rough mileage log is better than a guess.
• The longest day in that period and how often days above 60 per cent of a candidate range occur.
• Where the van is parked overnight, who has access, and whether a charger could be installed there.
• Annual mileage, current fuel and maintenance costs per mile, and the planned replacement date.
• Payload and towing needs, because some electric vans are rated for less payload or towing than their diesel equivalents; check the figure for each model you consider.

If your records are untidy, an assistant can help to standardise them. Paste a sample of the raw export and ask it to describe the columns, find gaps and suggest a clean layout; our tutorial on cleaning messy data covers that stage. Do not paste anything that identifies drivers beyond a van reference, and use a business-grade plan whose terms do not train on your content.

The worked example: six vans, one depot

An illustrative fleet of six vans runs from one yard that can take chargers. The assumptions below are placeholders, so replace every one with your own quote or tariff.

• Diesel van: 35 miles per gallon, fuel at $4.20 a gallon, so $0.12 a mile; maintenance $0.08 a mile. Total $0.20 a mile.
• Electric van: 2.8 miles per kilowatt-hour. Charging at the depot on a night tariff of $0.15 a kWh costs about $0.054 a mile; maintenance $0.05 a mile. Total about $0.104 a mile.
• Public rapid charging at $0.55 a kWh costs about $0.196 a mile before maintenance, so about $0.246 with it: more than diesel.
• Extra purchase cost of the electric van over the diesel equivalent, after expected resale value: $9,000. Charger and installation share per van: $1,500. Total extra: $10,500. Holding period: six years.

Saving per depot-charged mile is $0.20 minus $0.104, about $0.096. For a van to repay $10,500 in six years it needs $10,500 divided by (6 x $0.0964, or about $0.578) = about 18,150 miles a year, with no grant. With a confirmed grant of $3,000 per van the target drops to $7,500 divided by $0.578 = about 12,970 miles a year.

Now the sums for the three that matter, over six years. Van E: 17,500 x $0.096 x 6 = $10,080 saved against $10,500 extra, so $420 short without a grant and $2,580 ahead with $3,000. Van B: 13,750 x $0.096 x 6 = $7,920 saved, so $2,580 short without a grant and $420 ahead with it. Van C is blended: if 70 per cent of its 30,000 miles are depot-charged at $0.104 and 30 per cent use public chargers at $0.246, the average is about $0.147 a mile, a saving of $0.053 a mile. That is 30,000 x $0.053 x 6 = $9,540 against $10,500 extra: about $960 short without a grant, $2,040 ahead with one.

Notice what the figures say. At these prices the model does not shout "switch everything". It names Van E first, then B and C if the grant is real and the range plan works, and it says the other three should wait. Prices will differ for your fleet, so the lesson is the method: let mileage and charging decide the order.

Have AI build the sheet, then check it yourself

An assistant is useful here for structure: columns, formulas and sensitivity tables. It is unreliable on arithmetic across many cells and on prices it half remembers. So ask it for formulas, not answers, and put your inputs in one clearly marked block.

I run a fleet of six vans and want to decide which to replace with electric vans first. Build me a Google Sheets layout with:

1. An INPUTS block at the top (one cell per input, shaded yellow): diesel mpg, fuel price per gallon, diesel maintenance per mile, electric miles per kWh, depot tariff per kWh, public charging tariff per kWh, electric maintenance per mile, extra purchase cost, charger cost per van, holding period in years, grant per van.
2. A VANS table, one row per van: van ID, longest realistic day (miles), share of annual miles charged in public (%), annual miles, months to replacement, overnight charging (yes/no).
3. Calculated columns using formulas that reference only the INPUTS block: cost per mile diesel, cost per mile electric (blended by public share), annual saving, holding-period saving, extra cost, net with grant, net without grant, and a PASS/FAIL for range (longest day must be under 70% of an assumed range cell).
4. A RANK column ordering vans by net with grant, but only for vans that pass range and have overnight charging.

Show the formulas in plain text so I can check them. Do not fill in any numbers I have not given you.

When the sheet comes back, run two checks. First, type the six-van figures above into it and compare Van E with the hand sum ($10,080 saved, $420 short). If the sheet disagrees, find which formula differs rather than trusting either. Second, change one input, such as the depot tariff from $0.15 to $0.25, and see whether the ranking moves in the direction you expect.

Sensitivity: which assumption could flip the answer

The next job is finding which input the answer rests on. Ask the assistant to produce a table showing Van E and Van B's six-year result as the depot tariff moves between $0.10 and $0.30 a kWh, and again as the diesel price moves between $3.50 and $5.50 a gallon. A sample of what to expect, with the same assumptions:

The pattern: electricity price moves the result sharply, so a fixed-price night tariff, and not an average daytime one, matters. Before you commit, get a written tariff, and run the model on the price you can show, not the price you hope for.

Range is not one number

Quoted range is a test-cycle figure. Real range falls with cold weather, heavy loads, motorway speeds, heated cabs and battery ageing. Cold is the big one: published government cold-weather test figures show an electric vehicle's range can drop by 41 per cent at about minus 7 degrees Celsius (20 degrees Fahrenheit) with the cabin heater in use, with about two-thirds of the extra energy going on heating the cabin. That is why the filter above uses 70 per cent as a mild-weather planning figure, and why a fleet that works through freezing winters should plan nearer 55 to 60 per cent for those months, or use its own winter trial data. Preheating the cab while the van is still plugged in, which the same guidance recommends, takes part of that load off the battery. To make that concrete for Van C, suppose a manufacturer quotes 190 miles. At 70 per cent the planning figure is 133 miles. Van C's longest day is 150, so it fails the single-charge test and needs a mid-day top-up.

Ask the assistant to list the days in your data that exceed 133 miles. If Van C has four such days a quarter, the cheaper answer may be to reroute those four days to a diesel van and let Van C pass. If it has four a month, plan a public charge stop and add its cost to the model, which is what the 30 per cent public share did above. Time a real stop during the trial rather than assuming one, because charger speed, the van's own charging limit and queues all change how long it takes. Always check a van's real figures with a trial: borrow or hire an electric van of the model you are considering for two weeks and log the actual consumption.

A mistake that came from using the average day

Picture a small gardening firm that tested an electric van against its average daily mileage of 38 miles. The van passed with ease. Then, twice a month, the crew drove 120 miles to a large contract site, and the van arrived with its battery nearly empty. The fix came from a single column added to the sheet: longest day by month, not the average. When the firm re-ran the model, the van was rerouted to short jobs and a second van served the contract site, which turned a bad purchase into a sensible split.

The lesson is a rule for the sheet: any van whose longest day over 90 days exceeds your planning range fails until you decide how that day will be covered. Put the PASS/FAIL beside the cost, so a good saving never hides a bad route.

Charging access: depot, driver's home and public

The charging column decides more than the mileage column does, so record it in more detail than yes or no.

• Depot. Count the vans that will sleep there at once and the power available. Several chargers may need a larger electrical connection, which can take months and cost far more than the chargers themselves, so ask your electrician or the network operator early.
• Driver's home. Works for a van that goes home with the driver, if the driver has off-street parking, agrees to a charger, and you have a fair way to reimburse the electricity. The reimbursement rate is a tax and employment question, so check it with an adviser.
• Public only. Treat this as a last resort for everyday use, because rapid charging often costs more per mile than diesel, as the model above shows.

Ask the assistant to draft a one-page note for each depot or driver listing the practical questions: parking space, distance to the electrical panel, who owns the property and whether landlord consent is needed. You then verify each answer on site. Questions about consent and installation are for the landlord and a qualified electrician, not for an AI.

Grants: check them by hand, every time

Support for electric vans comes in several forms: grants towards the van, grants towards chargers, tax treatment and local schemes. These change often, have eligibility rules and sometimes close when the budget runs out. An assistant working from memory may describe a scheme that has ended or changed.

A safe process: use the assistant to draft your list of questions (eligible vehicle types, the date the van must be ordered or registered, whether the grant applies before or after tax, whether it can be combined with others, whether installation must use an approved installer). Then find each scheme on the funder's own official page, copy the rules and the date you read them into a notes cell beside the grant input, and keep a screenshot or a saved copy. Enter the grant in the model only when you have it confirmed in writing for your case. Keep two result columns, with and without, so you can see how much of the decision rests on it.

Turn the output into a switching order and timetable

Once the sheet ranks vans, ask the assistant to draft a plain timetable from your own data, and then edit it. With the six-van example it might read:

1. Now to month 3. Get a written night tariff, a depot power survey and a charger quote. Do a two-week trial of an electric van on Van E's route and log kWh per mile.
2. Month 3. Confirm or rule out the grant in writing. Decide on Van C: either plan the mid-day charge or move its longest days to a diesel van.
3. Months 6 to 8. Order the first electric van to replace Van E when its lease or finance ends.
4. Month 18. Repeat the model for Van B with fresh prices; do not assume the old ones hold.
5. Vans A, D and F. Review each at its replacement date, or sooner if parking or mileage changes.

A simple rule of thumb that fits the rest: do not build a five-year plan from today's prices. Plan the next vehicle properly and re-run the sheet when each replacement approaches. That is also kinder to your cash flow than ordering all at once.

One van, before and after the model

Van B shows why the sheet earns its keep. Before the model, the owner's note read: "Newish van, short routes, good candidate." After entering the data it read: longest day 85 miles (passes the 70 per cent test on a 133-mile planning range), depot charging yes, 13,750 miles a year, replacement due in 18 months, six-year net minus $2,580 without a grant and plus $420 with one. The decision changed from "switch now" to "switch at replacement if the grant is confirmed in writing". That is a calmer and cheaper decision than the first, and the sheet shows exactly which fact would change it.

Ask the drivers before you decide

Numbers do not cover everything. Drivers know which days run long, where vans really park, which customers have no easy parking for a charge stop and which jobs involve heavy loads. Sit with each driver for ten minutes with their van's row on a printout and ask what the sheet has wrong. Record the answers in a notes column. Often one comment, such as "on Fridays I do the far depot run", changes a van from pass to fail, and it never appeared in the telematics export because the run was logged under another van.

Keep the model honest after you start

Once the first electric van is working, record real figures: kWh used, miles driven, where it charged and what that cost. After 90 days replace the assumed miles per kWh with the measured number. If it is 2.2 instead of 2.8, the saving per mile falls and the whole ranking may shift. That is the real use of the sheet: it is a tool you update, not a verdict you file.

Review it with a second pair of eyes before you spend serious money; our tutorial on human review for AI work sets out a light way to do it. For other fleet jobs where AI helps with less risk, see seven fleet-manager jobs beyond reminders and the guide to AI route planning for small fleets, since better routes also change which vans pass the range filter. If you want help building and checking the model with your real figures, that is the kind of problem my AI implementation consultation is for, at the contact page.

Questions small fleets ask before switching vans

Which vans should a small fleet switch to electric first?

Usually the vans with high yearly mileage, predictable daily routes under about 70 per cent of the electric range (less if winters are hard), a place to charge overnight, and a replacement date coming up anyway. Vans that fail any of those four tests stay as they are until their circumstances change.

Can I trust an AI's cost figures for this?

Trust its structure, not its numbers. Let AI build the spreadsheet and the formulas, but type in your own prices, mileages and tariffs, and check two vans by hand. Assistants are weak at long chains of arithmetic and may quote old prices from memory.

Should I include grants in the model?

Show the model both ways: without any grant and with the grant you have confirmed in writing. Grants change often, carry conditions and sometimes run out, so check each one on the funder's own page and do not let a draft from an assistant be your source.

What if a van only occasionally does a very long day?

Model the longest realistic day, not the average. A van that does 40 miles on most days and 150 once a month may need a diesel van kept for that day, a public fast-charge stop planned into the route, or the job moved to another van.

Further reads

• Rental Car Companies Using AI: Seven Uses a Small Fleet Can Copy — Seven ways a small fleet can use AI beyond vehicle choice.
• Fleet Admin for Small Businesses: Automate Services, Checks and Fuel — Automate services, checks and fuel records for the vans you keep.
• How to Use AI for Scenario Planning: Best, Worst and Likely Cases — Best, worst and likely cases for any big spending decision.
• AI Software for Trucking Companies: What to Check for a Small Fleet — What to check when choosing fleet software.
• AI in Logistics for Small Delivery Firms: Routes, ETAs and Messages — Routes, ETAs and customer messages for small delivery firms.
• How to Find AI Savings Line by Line in Your Profit and Loss — Check the savings against real lines in your accounts.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: fueleconomy.gov cold-weather page (EV range and cabin heating), read 5 November 2026. No vendor prices are used: the model's prices, mileages and tariffs are illustrative and must be replaced with your own quotes, tariffs and funder pages.